Higher Rates of Decline for Women and Apolipoprotein E ε4 Carriers

This study assesses the risk of being female in addition to the well-known factors of age and apolipoprotein E ε4 status in the development and progression of Alzheimer disease based on longitudinal brain atrophy, cognitive decline, and CSF markers. APOE ε4 accelerated rates of decline, especially in women. The gender effect was at least as important as APOE ε4 status and showed weaker relationships to CSF markers. BACKGROUND AND PURPOSE: Age and the apolipoprotein E ε4 allele are well-known risk factors for Alzheimer disease, but whether female sex is also a risk factor remains controversial. It is also unclear how these risk factors affect rates of structural brain and clinical decline across the spectrum of preclinical to clinical Alzheimer disease. Our objective is to estimate the effects of apolipoprotein E ε4 and sex on age-specific rates of morphometric and clinical decline in late-onset sporadic Alzheimer disease. MATERIALS AND METHODS: With the use of linear mixed-effects models, we examined the effect of age, apolipoprotein E ε4, and sex on longitudinal brain atrophy and clinical decline among cognitively normal older individuals and individuals with mild cognitive impairment and Alzheimer disease (total = 688). We also evaluated the relationship between these effects and CSF biomarkers of Alzheimer disease pathology. RESULTS: Apolipoprotein E ε4 significantly accelerated rates of decline, and women in all cohorts had higher rates of decline than men. The magnitude of the sex effect on rates of decline was as large as those of ε4, yet their relationship to measures of CSF biomarkers were weaker. CONCLUSIONS: These results indicate that in addition to apolipoprotein E ε4 status, diagnostic and therapeutic strategies should take into account the effect of female sex on the Alzheimer disease process.

T he clinical presentation of Alzheimer disease (AD) is not uniform across individuals: in addition to atypical presentations 1,2 of AD, recent results show that the disease also presents differently in older compared with younger patients. 3,4 It is unclear, however, whether common genetic risk variants and sex also affect how the disease manifests and progresses.
In the United States, two-thirds of AD cases are women, 5 but because women live longer than men and older age is a known risk factor for AD, there remains controversy over whether women are at greater risk of development of AD than men. Several large epidemiology studies have found evidence of higher age-specific rates of incidence [6][7][8][9][10] and prevalence 11 of AD in women compared with men, though other studies have found no difference. 12,13 Elderly women, however, have higher amounts of AD pathology than elderly men, 14 and women with AD perform more poorly than men on cognitive assessment. 15 Assessing sex differences in age-specific cognitive and structural rates of decline may help elucidate this controversy.
The strongest known common genetic risk factor for sporadic AD is the apolipoprotein E (APOE) 4 allele. 16,17 APOE 4 increases the age-specific risk of development of AD in a dose-dependent manner 18,19 and lowers the age of onset. 18,20 Recently, we showed 3 that rates of both cognitive and structural decline decreased with age in individuals with mild cognitive impairment (MCI) and AD, but increased with age for the cognitively healthy elderly. Because 4 lowers the age of onset, age differences in rates of decline may have arisen partially from differences in 4 prevalence with age. Thus, to better understand AD biomarker trajectories, it is important to assess simultaneously the effects of 4 and age, as well as those of sex, on rates of clinical and structural decline.
We analyzed baseline and longitudinal data from cognitively healthy elderly (HC), MCI, and mild AD cohorts, age 65-90 years. We investigated the effects of 4 status and sex on cognitive and structural rates of change, and assessed whether such effects could be explained by baseline CSF concentrations of A␤  and the neurodegeneration-associated and phosphorylated 181p (p-) proteins.

Participants
We examined participants from the Alzheimer's Disease Neuroimaging Initiative (ADNI, www.adni-info.org). Participant enrollment criteria, MR image acquisition, and CSF collection and analysis methods are provided in the On-line Appendix.
We evaluated 688 participants, age Ն65 years at baseline, who had longitudinal cognitive evaluations: 211 HC, 333 patients with MCI, and 144 patients with AD. Of these, 188 HC, 273 patients with MCI, and 105 patients with AD also had longitudinal structural MR imaging data (Table 1). Longitudinal evaluations were performed at 6-or 12-month intervals for up to 24 (AD) or 36 (HC and MCI) months. The research protocol was approved by each local institutional review board, and written informed consent was obtained from each participant.

MR Image Processing
We quantified anatomical regional change in serial MR imaging with the use of Quarc. 21,22 We analyzed data from all available time points that passed local quality control (total ϭ 2244). Images that had degradation caused by motion, technical problems, significant clinical abnormalities (eg, hemispheric infarction), or changes in scanner vendor during the time series were excluded. 23 We examined rates of change in medial and lateral temporal lobe structures affected in early AD [24][25][26] and in whole-brain volume.

Genetic, CSF, and Clinical Measures
We grouped participants with respect to sex and APOE 4 status (none, 4 -, versus at least 1 4 allele, 4 ϩ ) ( Table 1 and On-line  Table 7). Baseline CSF data were available on approximately half of the ADNI participants. All participants were scored for Clinical Dementia Rating Scale, sum of boxes (CDR-SB), 27,28 cognitive subscale of the Alzheimer Disease Assessment Scale (ADAS-Cog), 29,30 and Mini-Mental State Examination (MMSE) 31 at each visit.

Mixed-Effects Modeling
Longitudinal cognitive and structural MR imaging atrophy outcomes (Y ij ) represent change with respect to baseline. This is expressed as the difference in test scores for cognitive measures and as a percentage of baseline size for cortical thickness change and region of interest volume change.
With the use of all available time points per participant, we investigated the dependence of atrophy rate and rate of clinical decline on 4 status and sex by use of a linear mixed-effects model, 32 controlling for baseline age, education, and, in the case of atrophy, baseline clinical severity. For each diagnostic group, the longitudinal outcome measurement Y ij at time t ij for participant i at follow-up time point j is Here, b 0 , b Cog , b Edu , b Age , b APOE , and b Sex are group regression parameters to be determined; C i , D i , A i , E i , and S i are covariates for participant i, respectively, mean-centered baseline clinical severity as measured by ADAS-Cog (for atrophic measures only: C i '0 when Y ij is a cognitive measure), mean-centered educational level (years of education), mean-centered baseline age, 4 status (E i ϭ 0 for 4 -, E i ϭ 1 for 4 ϩ ), and sex status (S i ϭ 0 for male, S i ϭ 1 for female); and ij is the within-participant error, assumed to be independent and identically normally distributed with zero mean and variance 2 . The first term on the right side of Eq. (1) incorporates mixed effects, allowing for different participant-specific rates of change: b 0 is the group fixed effect slope and ␤ 0i is the corresponding between-participant random effect slope, with zero mean, assumed to be normally distributed with variance 0 2 . Subsequent covariate terms involve fixed effects only.
We estimated the model parameters (including 0 and ) by use of the Matlab (R2009b) function nlmefit (MathWorks, Natick, Massachusetts). A follow-up set of analyses incorporated additional terms in Equation 1 for baseline CSF A␤ and p-concentrations to assess whether 4 or sex effects could be explained by CSF biomarker values.  Note:-N indicates number of participants. Values are mean (standard deviation, SD). MCIc ϭ MCI converters to AD. a MCI: 4 ϩ women are significantly younger than all other groups (all P Ͻ .01); 4 ϩ men are significantly younger than 4 Ϫ men and 4 Ϫ women (P Ͻ .01). b MCIc: 4 ϩ women are significantly younger than all other groups (all P Ͻ .05); 4 ϩ men are significantly younger than 4 Ϫ men and 4 Ϫ women (P Ͻ .05). c AD: 4 ϩ women are significantly younger than 4 Ϫ women (P Ͻ .05).

Rates of Decline in Healthy Controls
structures, with an additional 0.04%/year loss in the hippocampus, entorhinal cortex, and amygdala for each additional year of age above the group mean ( Table 2, b Age column). The presence of an 4 allele showed a large effect on annual rate of change in the same medial temporal regions, contributing an additional Ϫ0.42%/year loss in the hippocampus, Ϫ0.52%/year loss in the entorhinal cortex, and Ϫ0.63%/year loss in the amygdala ( Table  2, b APOE column). Sex significantly affected rate of change ( Table 2, b Sex column), with women showing higher rates of change than men for the hippocampus (an additional Ϫ0.25%/year), the entorhinal cortex (Ϫ0.49%/year), and the amygdala (Ϫ0.53%/year).
In contrast to the strong effects of 4 and sex on medial temporal atrophy rates, we did not find a significant association between these factors and rate of decline on any of the clinical measures in HCs.
The effects of age, 4, and sex on rates of decline in the entorhinal cortex and hippocampus are shown in Fig 1 for the HC, MCI, and AD cohorts, at the group average ages, educational levels, and ADAS-Cog scores. Fig 2 shows the effects of age, 4, and sex on rates of decline in CDR-SB and ADAS-Cog for the 3 cohorts, at the group average ages and educational levels. Table 3 shows the effects of age, 4 status, and sex on atrophy rates and rates of clinical decline in the MCI cohort. With the exception of the hippocampus and amygdala, increased age was associated with a slower rate of decline (b Age coefficients are positive) for all brain regions examined. Significant effects of 4 status were observed for all medial temporal lobe structures and for the inferior parietal cortex, with the additive effect of 4 on annual atrophy rate ranging from Ϫ0.28%/year to Ϫ0.94%/year. Independent of 4, sex significantly affected rate of change in all brain regions examined, except for the hippocampus: Women atrophied faster than did men, with the magnitude of the additive effect exceeding that of the 4 effect.

Rates of Decline in MCI
Significant 4 additive contributions to rates of cognitive decline were found for CDR-SB (0.38 points/year), ADAS-Cog (0.72 points/year), and MMSE (Ϫ0.81 points/year), whereas effects of female sex were significant for CDR-SB (0.26 points/year) and ADAS-Cog (1.40 points/year). Table 4 shows the effects of age, 4 status, and sex on rates of atrophy and clinical decline in AD participants. The effect of age on rates of change was significant for all brain regions examined, with increased age associated with lower rates of decline. The additive contribution to rate of decline for 4 was significant only for the amygdala (Ϫ0.91%/year) but showed a trend toward significance for the hippocampus and entorhinal cortex. Significant sex effects were found for all regions except for the hippocampus and amygdala, with women having higher rates of decline. There were

Effects of APOE 4 and Sex on Baseline CSF and Clinical Measures
Controlling for age and sex, 4 carriers showed significantly lower CSF A␤ concentrations than noncarriers, with the magnitude of the effect decreasing from HC to patients with MCI to those with AD (Fig 3 and On-line Table 5A). Relative to noncarriers, 4 carriers showed significantly higher CSF concentrations of and p-in the HC and MCI cohorts, but no significant differences were found for these biomarkers in the AD cohort.
Controlling for age and 4 status, there were no significant effects of sex on CSF A␤ or p-concentrations in any of the cohorts (Fig 3 and On-line Table 5A). For , the effect of sex approached significance for the MCI cohort only (P ϭ .060), with women showing higher concentrations than men.    Controlling for age and sex, performance on the clinical tests was significantly affected by 4 status in MCI participants only, with carriers showing worse performance than noncarriers for CDR-SB and ADAS-Cog, and showing a trend for worse performance on MMSE (On-line Table 5A and On-line Fig 1). Controlling for age and 4 status, no sex differences were found on the clinical tests in the patient cohorts, though MCI showed a trend toward significance for MMSE (P ϭ .072), with women performing more poorly.

Effects of Baseline CSF A␤ and p-on Rates of Decline
With A␤ in the model, significant effects of 4 and sex remained for MCI and AD participants, signifying that APOE 4 exerts an effect on atrophy rate in AD independent of its relation to A␤ (On-line Tables 3A-C). For HCs, however, there were no significant effects of 4 with CSF A␤ in the model. Adding an additional term for pconcentrations did not alter these results (On-line  Tables 4A-C), but this term was found to be significant in MCI for the amygdala, entorhinal cortex, ADAS-Cog, and MMSE, and in AD for the entorhinal cortex, rendering the A␤ term insignificant for all measures.

DISCUSSION
Our results show that changes in brain structure and function related to aging and AD do not progress uniformly across individuals but instead depend on age, sex, and APOE 4 status. Age differences in progressive atrophy and clinical decline, whereby older patients with MCI and AD decline at a slower rate than younger patients but older healthy adults decline at a faster rate than younger healthy adults, have been previously reported. 3,33 However, our finding that sex differences in atrophy rates are as large as differences associated with the well-known genetic risk factor, APOE 4, is novel, and has important implications for clinical practice, therapeutics research, and for advancing mechanistic understanding of AD.
The results showed that in all stages, from healthy aging through AD dementia, women had higher rates of brain atrophy than men, and the magnitude of the sex differences was at least as large as the magnitude of the APOE 4 effects. In HCs, sex differences were restricted to the medial temporal areas first affected in AD. In MCI and AD, the sex differences were more widespread, with weaker effects observed in medial temporal areas than in other brain regions. Additionally in MCI, in women compared with men, higher rates of atrophy were accompanied by higher rates of clinical decline.
These findings are consistent with prior large epidemiology studies 5-7,11,34 that showed higher rates of prevalence and incidence of AD in women than in men, with the differences between men and women comparable in magnitude to those between 4 carriers and noncarriers. They are also consistent with a recent meta-analysis that found lower cognitive performance for women than men diagnosed with AD. 15 A neuropathologic study 35 showed that women, especially if 4 carriers, are at higher risk of both neurofibrillary tangle (NFT) and amyloid plaque neuropathology than men in the earliest stages of AD (NFT stages I-III 26 ).
One possible explanation for the sex differences in HCs, in which women showed faster rates of atrophy in medial temporal areas, is that the HC women may be showing early signs of ADrelated neurodegeneration. However, the lack of sex differences in baseline CSF biomarkers of AD pathology in HCs does not support this view. The finding that CSF biomarkers did not explain the faster rates of decline occurring in women in any of the diagnostic groups suggests that other factors must be contributing to the sex differences. It has been argued that estrogens stimulate ␣-secretase activity and thus enhance nonamyloidogenic processing of amyloid-␤ precursor protein 36,37 ; the diminution in estrogen levels after menopause would then contribute to higher levels of AD pathology and poorer cognitive performance in women than in men. However, further research is needed to elucidate the basis of the observed sex differences.  Table 5A. *Significant differences.
The APOE 4 effects observed in the present study on longitudinal rates of change across cohorts are consistent with the elevated burdens of amyloid and pathology observed for 4 carriers compared with noncarriers at baseline. These baseline differences in CSF biomarkers between carriers and noncarriers agree with earlier reports 38,39 and with neuropathologic findings that 4 was associated with greater senile plaque and neurofibrillary tangle pathology in the elderly. 14 APOE 4 has further been associated with a higher plaque stage for a given age and allocortical NFT stage (Braak stages I-III, which correspond roughly with HC and early MCI) for 4 carriers compared with noncarriers, whereas at the later isocortical NFT stages (corresponding to late MCI and dementia), 4 gene dose was not an important predictor of pathology burden, 35,40 suggesting that 4 might exert its strongest effects in the prodromal stages of AD. Recently, Koffie et al 41 have shown that the 4 gene increases the amount of the synaptotoxic oligomeric A␤ in neuropil and its colocalization at synapses, even in nondemented control subjects, leading to synaptic injury and loss, a strong correlate of cognitive decline. 42 Our results showing elevated atrophy in 4 carriers generally, and our finding of marginally significant higher atrophy rates in predementia stages of AD for the medial orbito-frontal cortex 43 and inferior parietal lobule, sites of early amyloid deposition, 26 are consistent with these neuropathologic findings.
How 4 affects rates of cognitive decline across the preclinical, prodromal, and dementia stages of AD has been unclear, 20,44,45 but some studies have suggested that the effect of 4 is stronger in the earlier phases of the disorder. 39,46,47 Our results suggest that the accelerating effect of 4 on rates of decline diminishes with advancing disease stage, which comports with an earlier finding that 4 gene dose does not have a significant effect on the duration of AD, 20 and supports the hypothesis that as neurodegeneration advances, it becomes increasingly independent of initiating events. 48 This study has several limitations: The ADNI sample is not representative of the general population, and there was sex bias in MCI enrollment, with men outnumbering women. The HC and AD cohorts, however, showed more balanced sex representation. Because similar sex effects were observed across groups, they are unlikely to have arisen from enrollment bias. There is insufficient information within ADNI to address issues of whether history of hormone replacement therapy or number of years since menopause may have influenced the observed sex differences. Finally, statistical power was limited with respect to analyses of CSF biomarker data. Larger population-based studies that can systematically address hormonal issues, and other medical issues that may differ between the sexes, are needed to elucidate the basis of the observed sex differences in rate of atrophy and cognitive decline.

CONCLUSIONS
Our results show that women and APOE 4 carriers in ADNI have higher rates of decline in normal aging, MCI, and AD, and that these effects are not fully explained by baseline CSF concentrations of AD-related proteins. Because two-thirds of AD cases in the United States are women, and because the higher rates of decline in women compared with men were at least as large as those related to the major genetic risk factor, APOE 4, it is of particular importance that sex differences in rates of decline in aging and AD be taken into account in the clinical setting and in therapeutics research. Greater understanding of the mechanistic basis of these differences likely will facilitate further understanding of AD etiology.
Disclosures: Dominic Holland-RELATED: Grant: NIH NIA,* Comments: This research was supported by NIH grants R01AG031224, R01AG22381, U54NS056883, P50NS22343, and P50MH081755 (A.M.D.); NIA K01AG029218 (LKM); NIH-NIBIB T32 EB005970 (RSD); UNRELATED: Grants/Grants Pending: NIH*; Patents (planned, pending or issued): UCSD,* Comments: A patent application for Quarc, a method to quantify anatomical regional change developed at UCSD, has been filed through the UCSD Technology Transfer Office ("Longitudinal registration of anatomy in magnetic resonance imaging"; Application number: 12/742,642; Publication number: US 2010/0259263 A1). Rahul Desikan-RELATED: Grant: T32 EB005970,* Comments: NIH Training Grant. Anders Dale-RELATED: Grant: UCSD,* Comments: Numerous NIH grants as PI and Co-I; UNRELATED: Grants/Grants Pending: UCSD*; Patents (planned, pending or issued): UCSD,* MGH,* Comments: I am an inventor of a provisional patent on Restriction Spectrum Imaging, and other issued patents in the areas of MRI acquisition and post-processing; Stock/Stock Options: CorTechs Labs Inc, Comments: I am a founder and hold equity in CorTechs Labs, Inc, and also serve on its Scientific Advisory Board. The terms of this arrangement have been reviewed and approved by UCSD in accordance with its conflict-of-interest policies; Other: UCSD* Comments: I am the principal investigator of a research agreement between UCSD and General Electric Medical Systems. Linda McEvoy-RELATED: Grant: National Institute on Aging,* Comments: Grant support from K01AG029218, R01AG22381; Support for Travel to Meetings for the Study or Other Purposes: National Institute on Aging,* Comments: Grants listed above used to support presentation of data at 2012 Alzheimer's Association International conference (*money paid to institution).
A.M.D. is a founder and holds equity in CorTechs Labs, Inc, and also serves on its Scientific Advisory Board. The terms of this arrangement have been reviewed and approved by the University of California, San Diego, in accordance with its conflict of interest policies. McEvoy's spouse is President of CorTechs Labs, Inc.
Data used in preparation of this article were obtained from the ADNI database (adni.loni.ucla.edu). As such, the investigators within the ADNI contributed to the design and implementation of ADNI and/or provided data but did not participate in analysis or writing of this report. A complete listing of ADNI investigators can be found at: http://adni.loni.ucla.edu/wp-content/uploads/how_to_apply/ADNI_